20

I'm a first-year PhD student in a computer science department. As is usual at this school, my first three quarters are spent rotating with different professors, finding an adviser and a research fit that I like. However, unlike previous rotations, I can see myself working in this subfield for my PhD, and I've moved from wondering what subfield I'll be working in, to wondering how to get acclimated in this field.

I've always been a person to run an "outer loop" of self-reflection, advice-asking, and habit-forming. In undergrad, I followed a few blogs about making the most of college. However, it's been a lot harder for me to find useful advice about doing top-quality work in getting a PhD.

I hear that in the first part of your PhD, students are significantly less "productive" than in the fourth year and onward. That makes sense, and I'm becoming comfortable with the banging-head-against-wall feeling that is creeping up on me. However, I'd like to know what the three years of "unproductive" time teaches you to do, so I know what I should focus my energy on during these years.

I've heard it's important to read papers, develop a "taste" for useful problems, and hone in on a larger research question. However, each of those has a lot of follow-up questions that few people seem to talk about. Reading papers: how many? what about? what for? Taste: how do I get as many "data points" as possible in learning what's a useful problem?

Edit: I'm at school in the United States, and I'm unsure whether I'd want to go to academia, industry or run a startup. If I were to bet right now, I'd say 50% academia, 30% startup, 20% industry.

  • 4
    These all depends entirely on the student, field, and problem. Spend that time figuring out what works for you. – Austin Henley Mar 28 '17 at 22:52
  • I feel like there are common reasons for reading papers, though - are you doing it to learn about the field in general? to look for innovations you can carry into your research? to solve a particular problem you have in your current research project? to see the next "hot" topics you should tackle for the year? – Mark Miller Mar 28 '17 at 23:06
  • Yes, that is correct. – Austin Henley Mar 28 '17 at 23:09
  • And whether those questions are strong or silly depend on your own field, self, and problem? – Mark Miller Mar 28 '17 at 23:11
  • Doing all of them could prove helpful, maybe. – Austin Henley Mar 28 '17 at 23:13
14

To get the ball rolling, I will offer an answer. I will invite others to edit my answer to improve it, or write another answer.

  • In the early part, spend about 5 - 10% of your time looking at journal articles and other scientific literature (if you are on top of your coursework).

  • Form one or more study groups, if for no other reason than to get practice communicating about your subject area.

  • Attend seminars.

  • Visit office hours.

  • Try to keep your notes organized.

  • If there are foundational exams in your program, get some old exams and start looking at them, as time permits.

  • Find fun ways of getting exercise.

  • 3
    Why specifically journal articles? For most parts of CS I believe conference papers are (also) of high quality. Also, this answer seems to assume a USA style PhD (which I suspect is what the OP is after), it may be good to emphasise this for other readers. – J. Doe Mar 29 '17 at 11:36
  • 3
    @J.Doe - I didn't mean to restrict the reading to journals as opposed to proceedings. I meant scholarly or technical articles. What term would you use? – aparente001 Mar 29 '17 at 14:12
  • Perhaps "scientific articles" or "scientific literature" could work? – J. Doe Mar 30 '17 at 11:56
  • To elaborate on the second point, practice communicating your ideas is particularly important in academia. Try to seek a variety of formal and informal opportunities to explain existing literature, your hypotheses, and the result and conclusions from any data you collect. Make sure you can speak to different audiences as well; you should be able to adjust your explanations so they make sense for experts in your field, researchers outside your field, and individuals without higher education. – Adam Bosen Mar 30 '17 at 21:22
  • @J.Doe this question is not about CS. Journal is totally fine. – Rüdiger Mar 31 '17 at 19:20
10

In addition to @aparente001's excellent advice, I would suggest:

  • Establish a 'keyword' list of research terms as you read through the relevant papers. This becomes important as it will help you refine your research as time progresses.

Other advice, based on my own experience:

  • Be prepared to collaborate with your supervisor to write/publish papers based on your work (I did this, published 5 articles before graduation).
  • Allow some time for other interests (e.g. music)
  • Create a physical workspace (or several).
  • Make time for family and friends (this is a very important factor as they are often your support).
  • maintain proper sleeping patterns (time and duration)
  • 3
    Great additions. I especially like the sleeping pattern advice. – aparente001 Mar 29 '17 at 14:15
4

I've heard it's important to read papers, develop a "taste" for useful problems, and hone in on a larger research question. However, each of those has a lot of follow-up questions that few people seem to talk about.

The reason why those questions are not typically answered is because they are (a) very individual and (b) impossible to put a concrete number on. I would argue that focusing on "how much I still need to read" is not the right question to ask anyway.

Reading papers: how many?

Until you know what the common research directions are, and what the state of the art and open problems are in the one(s) that interest you. Frankly, most students seem to typically stop primarily reading and move on to primarily doing when they get bored reading because they feel they know most of the important stuff anyway.

Generally speaking, I tell my students that it's better to read 1 good paper than 5 mediocre ones. I would also initially not recommend taking the most complicated paper you can find and trying to understand everything. Initially, breath is more important than trying to "get" every little detail.

I also suggest that you start thinking about what you could see yourself doing as early as possible. Do you see any follow-up questions, and do you have an idea how they could be answered? Do you see yourself conducting a similar study to the one explained in the paper? What do you still need to learn to do such a study?

what about?

Initially: very broad. As soon as you get a feeling for what kind of papers are of particular interest to you: those.

what for?

For three reasons: (1) to learn what scientifically the state of the art is, (2) to learn what research methods are commonly used to address which problems in the field (and, implicitly, what the typical expectations in terms of scientific rigour are, e.g., related to sample sizes), and (3) to learn how to write up and sell your research in your community.

Taste: how do I get as many "data points" as possible in learning what's a useful problem?

Primarily by reading broadly. I am not sure what else you can do.

  • Reading broadly and reading one good paper for complete understanding sound a bit at odds with each other. End result, this could exacerbate school anxiety, if the student tended in that direction. I'm saying this because I don't think you intended your advice to have that effect. – aparente001 Mar 29 '17 at 14:14
3

I will offer one additional piece of advice that I think is important:

  • Set up and tune to your liking several search alerts that notify you of interesting new papers in the literature.

This is essentially the only scalable way of keeping up with research today. There are many flavors of search alerts, including

  • Journal TOC alerts: gives you a title & authors list of every paper in the latest issue of a specific journal. Only useful for the key journals in your field.
  • Preprint alerts: e.g. on the arXiv you can set up daily alerts with new papers in the field you are interested in, these are title/authors/abstract lists.
  • Topical search alerts: e.g. on ScienceDirect you can set up search alerts for custom keywords. You can use this to track papers on a specific topic, papers by specific authors, etc.
  • Non-paper-based: e.g. mailing lists for specific research interests/software/etc. that you want to follow, RSS feeds of blogs by scientists in your field, etc.

You probably will want to start by adding many of these, and then unsubscribe from the ones you find less useful. And your interests will probably change over time.

  • Excellent advice on the search alerts - I found them invaluable! – user70612 Mar 31 '17 at 8:59
  • @semi-extrinsic: Wait, arXiv has an abstract-search subscription service? I whipped one up myself this past week because I only saw the subject-matter [e.g, cs.HC] subscription service – Mark Miller Mar 31 '17 at 18:19
  • This helpful information is a bit buried here. I suggest you write a question keyed to this answer. – aparente001 Mar 31 '17 at 19:58
  • @aparente001 what information / who are you directing this comment to? – Mark Miller Mar 31 '17 at 23:16
  • @MarkMiller I was referring to the subject-matter search, by "title/authors/abstract/ list" I meant these are the items in the email you get. – semi-extrinsic Apr 1 '17 at 16:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.